User Representation Learning for Social Networks: An Empirical Study

Gathering useful insights from social media data has gained great interest over the recent years. User representation can be a key task in mining publicly available user-generated rich content offered by platforms. The way to automatically create meaningful observations about users of network is obtain real-valued vectors for with user embedding learning models. In this study, we presented one most comprehensive studies literature terms high-quality representations leveraging state-of-the-art text approaches. We proposed novel doc2vec-based method, which encode both textual and non-textual information into low dimensional vector. addition, various experiments were performed investigating performance techniques concepts including word2vec, doc2vec, Glove, NumberBatch, FastText, BERT, ELMO, TF-IDF. also shared new dataset comprising 500 manually selected Twitter five predefined groups. contains different activity such as comment, retweet, like, location, well actual tweets composed users.